jclark@sdcc6.ucsd.edu (John Clark) (04/03/91)
Today, 2-Apr-91, I posted a request regarding the SPEC benchmarks and how to run them on small kernel based CPU's. Also, what to do about the FORTRAN benchmarks. Several replies were to get the f2c converter and then do the usual compilation and run. The next question is, has someone run this converted code against say the VAX/VMS FORTRAN to get a comparison of the 'efficiency' of the conversion. For instance, I convert then compile, the 'competition' compiles directly. I lose because, REAL's are passed by reference or REAL*4 are used but are 'promoted' in the converted stuff, or some other 'efficiency' that has been built into the direct FORTRAN compiler (complex arithmetic operations could be implement directly or via subroutine calls, etc.) Have others had this problem and gotten some numbers for comparison? This would solve part of my problem, in that the result could run on the BSD os I have. It does not solve the 'real time' aspect. In that regard I would like to know if there are others who have considered, vxWorks, vrtx, os9, psos, mach (without unix server support), or others, for small-kernel OS's. (Small being mostly in the mind of the beholder, but basically less than a unix port but more than naked machine.) -- John Clark jclark@ucsd.edu